In [7]:
%pylab inline
from IPython.display import display, HTML, Audio, Image
from scipy.io import wavfile
from nvx.dsp import *
import seaborn as sns
sns.set_style("whitegrid")
sns.set_palette("Set2")
figsize(12,5) 
Populating the interactive namespace from numpy and matplotlib

In the audio office

Low reverberant environment, but still, the voice sounds closer:

In [2]:
lieu = 'bureau'
d = {}
fe,s_p = wavfile.read('aec_rec/BU_triby/'+lieu+'_triby_preprocess.wav')
fe,s_p_d = wavfile.read('aec_rec/BU_triby/'+lieu+'_triby_preprocess_derev.wav')
d['Without dereverberation'] = s_p*4
d['With dereverberation'] = s_p_d*4

display_wavform_db_dict(d,['Without dereverberation','With dereverberation'],'In the '+lieu,mindb=-10,style='-')
d['Without dereverberation'],d['With dereverberation'] = s_p,s_p_d
display_audio_dict(d,['Without dereverberation','With dereverberation'],lieu,'In the '+lieu)
Without dereverberation:
With dereverberation:

At the cafeteria

Higher reverberant environment, the voice sounds cleaner and closer:

In [3]:
lieu = 'cuisine'
d = {}
fe,s_p = wavfile.read('aec_rec/BU_triby/'+lieu+'_triby_preprocess.wav')
fe,s_p_d = wavfile.read('aec_rec/BU_triby/'+lieu+'_triby_preprocess_derev.wav')
d['Without dereverberation'] = s_p*4
d['With dereverberation'] = s_p_d*4

display_wavform_db_dict(d,['Without dereverberation','With dereverberation'],'In the '+lieu,mindb=-10,style='-')
d['Without dereverberation'],d['With dereverberation'] = s_p,s_p_d
display_audio_dict(d,['Without dereverberation','With dereverberation'],lieu,'In the '+lieu)
Without dereverberation:
With dereverberation:

In the meeting room

Higher reverberant environment, again the voice sounds closer, still no distortion:

In [4]:
lieu = 'reunion'
d = {}
fe,s_p = wavfile.read('aec_rec/BU_triby/'+lieu+'_triby_preprocess.wav')
fe,s_p_d = wavfile.read('aec_rec/BU_triby/'+lieu+'_triby_preprocess_derev.wav')
d['Without dereverberation'] = s_p*4
d['With dereverberation'] = s_p_d*4

display_wavform_db_dict(d,['Without dereverberation','With dereverberation'],'In the '+lieu,mindb=-10,style='-')
d['Without dereverberation'],d['With dereverberation'] = s_p,s_p_d
display_audio_dict(d,['Without dereverberation','With dereverberation'],lieu,'In the '+lieu)
Without dereverberation:
With dereverberation:

In the bathroom

Highly reverberant environment, close to the one obtained in a kitchen. We hear a little distortion but intelligibility is significally improved:

In [5]:
lieu = 'sdb'
d = {}
fe,s_p = wavfile.read('aec_rec/BU_triby/'+lieu+'_triby_preprocess.wav')
fe,s_p_d = wavfile.read('aec_rec/BU_triby/'+lieu+'_triby_preprocess_derev.wav')
d['Without dereverberation'] = s_p*4
d['With dereverberation'] = s_p_d*4

display_wavform_db_dict(d,['Without dereverberation','With dereverberation'],'In the '+lieu,mindb=-10,style='-')
d['Without dereverberation'],d['With dereverberation'] = s_p,s_p_d
display_audio_dict(d,['Without dereverberation','With dereverberation'],lieu,'In the '+lieu)
Without dereverberation:
With dereverberation:

In the emergency stairs

The worst reverberant environment, without dereverberation we barely understand what I'm saying. After dereverberation we hear distortion in the speech signal... but we understand what is said:

In [6]:
lieu = 'escalier'
d = {}
fe,s_p = wavfile.read('aec_rec/BU_triby/'+lieu+'_triby_preprocess.wav')
fe,s_p_d = wavfile.read('aec_rec/BU_triby/'+lieu+'_triby_preprocess_derev.wav')
d['Without dereverberation'] = s_p*4
d['With dereverberation'] = s_p_d*4

display_wavform_db_dict(d,['Without dereverberation','With dereverberation'],'In the '+lieu,mindb=-10,style='-')
d['Without dereverberation'],d['With dereverberation'] = s_p,s_p_d
display_audio_dict(d,['Without dereverberation','With dereverberation'],lieu,'In the '+lieu)
Without dereverberation:
With dereverberation: